Children have been found to learn less in school when there are more boys in the classroom. We make two contributions to the literature. First, we revisit the empirical approach based on within-school grade-level variation. We find that with grade retention (a feature of most school systems) the approach is subject to a selection bias that generates effects of boys on their classmates‘ skills in the absence of any peer effects. We propose an alternative based on variation across enrollment cohorts. Second, we implement both approaches using data on around 3.6 million children in 14,000 primary schools in Texas, North Carolina, and Italy. We find adverse effects of boys on how much children learn in school using both the grade-level and the enrollment-cohort approach. The enrollment-cohort approach yields adverse effects on boys that are at least double the adverse effects on girls in all three school systems. By contrast, consistent with the selection bias we highlight, the grade-level approach yields similar adverse effects on boys and girls in the school systems with substantial grade retention.